As semiconductor manufacturing advances to ultra-large scale integration (ULSI), the devices on semiconductor wafers shrink to sub-micron dimension and the circuit density increases to several million transistors per die. In order to accomplish this high device packing density, smaller and smaller feature sizes are required. This may include the width and spacing of interconnecting lines and the surface geometry such as corners and edges, of various features.
The requirement of small feature sizes with close spacing between adjacent features requires high resolution photolithographic processes. In general, photolithography utilizes a beam of light, such as UV waves, to transfer a pattern from a photolithographic mask onto a photoresist coating through an imaging lens. The mask includes opaque and transparent parent regions such that the shapes match those of the openings in the resist coating in the desired or predetermined pattern.
One technique currently being investigated for improving the resolution of the photolithographic process is known as phase shift lithography. With phase shift lithography the interference of light rays is used to overcome diffraction and improve the resolution and depth of optical images projected onto a target. In phase shift lithography, the phase of an exposure light at the object is controlled such that adjacent bright areas are formed preferably 180.degree. out of phase with one another. Dark regions are thus produced between the bright areas by destructive interference even when diffraction would otherwise cause these areas to be lit. This technique improves total resolution at the object and allows resolutions as fine as 0.25 .mu.m to occur.
An early patent in this field, U.S. Pat. No. 4,360,586 to Flanders et al, was issued on Nov. 23, 1982 and assigned to MIT. This patent was directed to exposing periodic optical features on an object surface. The features were characterized by a spatial period p. According to the invention, a source of radiant energy of wavelength .lambda. illuminates a surface to be exposed through a mask having a spatial period separated from the surface by a distance approximately S.sub.n =p.sup.2 /n.lambda., where n is an integer greater than one.
With respect to semiconductor fabrication numerous laboratory techniques have been proposed to employ phase shifting in the photopatterning of semiconductor wafers. Most of the work in this area has centered around either "Alternating Phase Shifting", "Subresolution Phase Shifting", or "Rim Phase Shifting" experiments In general, in each of these techniques a phase shift mask or reticle is constructed in repetitive patterns of three distinct layers of material. An opaque layer on the mask provides areas that allow no light transmission therethrough, a transparent layer provides areas which allow close to 100% of light to pass through and a phase shifter layer provides areas which allow close to 100% of light to pass through but phase shifted 180.degree. from the light passing through the transparent areas. The transparent areas and phase shifting areas are situated such that light rays diffracted through each area is canceled out in a darkened area therebetween. This creates the pattern of dark and bright areas which can be used to clearly delineate features of a pattern on a photopatterned wafer.
"Alternating Phase Shifting" as disclosed in [1] is a spatial frequency reduction concept similar to the method disclosed in the Flanders et al patent. It is characterized by a pattern of features alternately covered by a phase shifting layer. "Subresolution Phase Shifting" [2] promotes edge intensity cut off by placing a subresolution feature adjacent to a primary image and covering it with a phase shifting layer. "Rim Phase Shifting" [3] overhangs a phase shifter over a chrome mask pattern.
In general, these phase shifting techniques have not been adapted to large scale semiconductor manufacturing processes. One problem with applying phase shifting lithography into practical use, in manufacturing semiconductors, is the difficulty in reticle mask making, inspection, and repair. The process must be compatible with manufacturing conditions, (i.e. inexpensive, repetitive, clean) and prior art laboratory techniques have not heretofore met these criteria.
A representative state of the art semiconductor laboratory process for making a phase shift mask or reticle is disclosed in reference [4]. This process was also generally disclosed in the Flanders et al patent. This process is shown in FIGS. 1A-1C and is termed a "lift off process".
The "lift off process" may be used to fabricate a reticle on hard copy of an individual drawing for a semiconductor circuit layout. The reticle may then be used directly as a mask in the photopatterning process or may be used to produce a photomask. As an example, this reticle may be used to pattern a wafer surface in a stepped pattern transfer. DRAM's and SRAM's because of their repetitive nature are adapted to manufacture in this manner.
Referring to FIG. 1A, with the "lift off process" a transparent quartz substrate 10 has a film of an opaque material such as chromium (CR) patterned thereon. The chromium (CR) may be deposited and patterned onto the substrate 10 by a conventional process such as electron beam deposition and photolithography. In the example of FIG. 1A, the pattern is a periodic arrangement of chromium (CR) light blockers 12 and spaces 14 patterned on the quartz substrate 10.
A layer of resist 16 is then deposited and patterned over the patterned chromium (CR) light blockers 12 and spaces 14. Every other space 14 is covered with resist 16 such that an alternating pattern of phase shifters and openings will be ultimately formed. As shown in FIG. 1B the resist 16 is patterned in a straight wall profiles such that a subsequent etching process aids the "lift off" step.
With reference to FIG. 1B, after deposition of the resist 16, a film of phase shifter material such as (SiO.sub.2) is blanket deposited over the photoresist 16 and patterned openings 14. The phase shift ultimately obtained is a function of the thickness "T" and refractive index of this phase shifter material, which are preferably selected to provide a 180.degree. phase shift.
As shown in FIG. 1(C) the phase shifter material (SiO.sub.2) is then "lifted off" with the remaining layer of photoresist 16 by stripping or etching away the photoresist 16. This leaves a phase shifter 18 in every other opening 14 between the chromium (CR) light blockers 12. This provides an alternating phase shifting pattern as previously explained.
A problem with this "lift off process" is that it is a defect prone, inconsistent, messy procedure not suitable for large scale manufacturing. Large chunks of (SiO.sub.2) material are lifted by the etching process and are difficult to remove from the finished reticle. These contaminants may cause subsequent patterning errors of the finished wafer. Another problem with the "lift off process" is that it is difficult to accurately control the thickness of the phase shifters 18.
In the finished reticle each phase shifter 18 preferably has a thickness "T" (FIG. 1C) that produces a 180.degree. phase shift for light passing therethrough. This optimal thickness can be determined by the formula ##EQU1## where T=thickness of phase shift material
i=an odd integer PA1 .lambda.=wavelength of exposure light PA1 n=refractive index of phase shifter material at the exposure wavelength
Any deviation from this optimal thickness adversely affects the phase shift ultimately obtained. An "edge effect" may occur, for example, at the edge of a phase shifter 18 deposited over the edge of a chromium (CR) light blocker 12. This may be due to the conformal deposition of the phase shifter material over the chromium (CR) light blocker 12 which causes the phase shifter to be slightly thicker at the edges.
Another prior art method for forming a phase shift reticle is shown in FIGS. 2A-2C. As before, a pattern of opaque light blockers 20 such as chromium (CR) is formed on a quartz substrate 22. As shown in FIG. 2B, photoresist 24 is then patterned on the chromium light blockers 20 in every other opening 25. As shown in FIG. 2C openings 26 not covered with photoresist are then etched to form phase shifter openings 28 etched back to a depth "d" into the substrate 22. Light passing through an etched back phase shifter opening 28 is phase shifted relative to light passing through an adjacent transmission opening 30, which must travel through the full thickness of the quartz substrate.
This prior art reticle fabrication method is also subject to several disadvantages. Firstly, the etch depth "d" into the quartz 22 requires a highly controlled, uniform etch over a relatively large (example--5") reticle surface. Since it is necessary to phase shift within a margin of .+-.10.degree. or less this translates to an etch tolerance of about 5%. This is difficult to achieve under manufacturing conditions.
Further the prior art etch back technique shown in FIGS. 2A-2C causes a finite loss in light transmission due to the roughness of the exterior surface of the etched back phase shifter openings 28. Another disadvantage of this etch back technique is the difficult requirement of having to accurately pattern photoresist 24 over the topology of the opaque light blockers 20.
The process of the present invention is directed to a novel process that overcomes these prior art limitations. The process of the invention provides a clean, repetitive, technique for forming accurate phase shifting reticles suitable for large scale semiconductor manufacturing. Moreover, the process of the invention can be used to provide phase shifters having a smooth and optically flawless surface.